Goodness-of-fit tests for the exponential distribution based on U-empirical kernel density estimators

Abstract

Characterizations based on the equidistribution of two statistics have become very popular for the construction of goodness-of-fit tests. Various approaches that use different U and V-empirical functions have been proposed. In this paper, we suggest a new method based on U-empirical kernel density estimators. We propose a class of exponentiality tests based on recent characterizations from [1] and [2]. We examine their properties using simulated powers. Their isotones will also be presented for some choice of alternative distributions.